OpenAI is rolling out a new safety feature for ChatGPT: 'Trusted Contact.' Users can now designate a friend or family member to be alerted if the AI detects conversations indicating a serious risk of self-harm.
To set it up, the user invites a trusted person who must accept. When ChatGPT's automated system flags a concerning conversation, it first encourages the user to reach out themselves. If a specially trained human review team confirms the severity, ChatGPT sends a general notification to the trusted contact via email, text, or in-app alert-without sharing chat transcripts or summaries.
OpenAI says the feature was developed with guidance from clinicians and mental health organizations. The company has faced scrutiny after past incidents involving users and self-harm. The feature is voluntary and designed to help loved ones check in during critical moments.